SERIES

IZA DP No. 10398

PAPER

The Effects of Computers on Children’s Social

Development and School Participation:
Evidence from a Randomized Control Experiment
DISCUSSION

Robert W. Fairlie
Ariel Kalil

November 2016

Forschungsinstitut
zur Zukunft der Arbeit
Institute for the Study
of Labor
The Effects of Computers on Children’s
Social Development and School
Participation: Evidence from a
Randomized Control Experiment

Robert W. Fairlie
University of California, Santa Cruz,
NBER and IZA

Ariel Kalil
University of Chicago

Discussion Paper No. 10398

November 2016

IZA

P.O. Box 7240

53072 Bonn
Germany

Phone: +49-228-3894-0
Fax: +49-228-3894-180
E-mail: iza@iza.org

Any opinions expressed here are those of the author(s) and not those of IZA. Research published in
this series may include views on policy, but the institute itself takes no institutional policy positions.
The IZA research network is committed to the IZA Guiding Principles of Research Integrity.

The Institute for the Study of Labor (IZA) in Bonn is a local and virtual international research center
and a place of communication between science, politics and business. IZA is an independent nonprofit
organization supported by Deutsche Post Foundation. The center is associated with the University of
Bonn and offers a stimulating research environment through its international network, workshops and
conferences, data service, project support, research visits and doctoral program. IZA engages in (i)
original and internationally competitive research in all fields of labor economics, (ii) development of
policy concepts, and (iii) dissemination of research results and concepts to the interested public.

IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion.
Citation of such a paper should account for its provisional character. A revised version may be
available directly from the author.
IZA Discussion Paper No. 10398
November 2016

ABSTRACT

The Effects of Computers on Children’s

Social Development and School Participation:
Evidence from a Randomized Control Experiment*

Concerns over the perceived negative impacts of computers on social development among
children are prevalent but largely uninformed by plausibly causal evidence. We provide the
first test of this hypothesis using a large-scale randomized control experiment in which more
than one thousand children attending grades 6-10 across 15 different schools and 5 school
districts in California were randomly given computers to use at home. Children in the
treatment group are more likely to report having a social networking site, but also report
spending more time communicating with their friends and interacting with their friends in
person. There is no evidence that computer ownership displaces participation in after-school
activities such as sports teams or clubs or reduces school participation and engagement.

JEL Classification: I20

Keywords: computers, ICT, education, social development, school participation, experiment

Corresponding author:

Robert W. Fairlie
Department of Economics
University of California, Santa Cruz
Santa Cruz, CA 95064
USA
E-mail: rfairlie@ucsc.edu

* We thank Computers for Classrooms, Inc., the ZeroDivide Foundation, and the NET Institute for
generous funding for the project. We thank seminar participants at the IZA Conference on Risky
Behaviors, CESifo Institute, and Teachers College at Columbia University, and Ben Hansen for helpful
comments and suggestions. We thank the many school teachers, principals and administrators that
helped run the computer giveaway program, which include Jennifer Bevers, Bruce Besnard, John
Bohannon, Linda Coleman, Reg Govan, Rebecka Hagerty, Kathleen Hannah-Chambas, Brian Gault,
David Jansen, Cynthia Kampf, Gina Lanphier, Linda Leonard, Kurt Madden, Lee McPeak, Stephen
Morris, Joanne Parsley, Richard Pascual, Jeanette Sanchez, Zenae Scott, Tom Sharp, and many
others. We thank Shilpa Aggarwal, Julian Caballero, David Castaneda, James Chiu, Samantha
Grunberg, Keith Henwood, Cody Kennedy, Nicole Mendoza, Nick Parker, Miranda Schirmer, Glen
Wolf and Heidi Wu for research assistance. Finally, special thanks go to Pat Furr for donating many
computers for the study and distributing computers to schools.
I. Introduction

Computer use among children has risen steadily over time and so too have concerns over

its negative effects on social development.1 Concerns center on the substitution of actual

interpersonal interactions with peers and parents for virtual ones, computer use’s displacement of

time on “developmentally meritorious” activities (e.g., sports and social activities), and the

exposure to cyberbullying or other inappropriate digital media content (Subrahmanyam, Kraut,

Greenfield & Gross, 2001). For instance, increased computer use could displace putatively more

meaningful face-to-face social contact and weaken social bonds with family and close friends

(Kraut et al., 1998).2 In addition, time spent surfing the web, on social networking sites, or

playing computer games is by definition time that cannot be allocated to sports, clubs, or other

extracurricular activities. Participation in extracurricular activities is seen as a key forum for

maintaining existing friendships and developing new ones (Crosnoe, 2011; Schaefer et al. 2011;

Eccles & Templeton, 2002). At its extreme, social isolation may increase children’s avoidance of

school through absences or chronic tardiness.

On the other hand, children's "alone time" on computers may extend social relationships

by connecting with others through virtual interactions and actually facilitate subsequent

interactions in person. In this way, computer ownership may complement, rather than substitute

for, interpersonal interactions. Computers are used extensively for communication and social

networking by children (Kaiser Family Foundation 2010; and Pew Internet Project 2013). The

1
Currently 86 percent of children have access to a computer at home, though levels of access are much
lower among minority and low-income children. See Bureau of Labor Statistics (2014); National
Telecommunications and Information Administration (2013). Also, see Hoffman and Novak (1998);
Servon (2002); Mossberger, Tolbert, and Stansbury (2003); Ono and Zavodny (2007); Fairlie (2004);
Goldfarb and Prince (2008) for earlier examples of analyses of disparities in computer access.
2
Similar concerns were expressed earlier over television crowding out more productive activities (see
Zavodny 2006 for example).

2
net effects of these opposing forces is unknown (Putnam 2000; Bauernschuster, Falck and

Woessmann 2014).

Surprisingly, very few previous studies rigorously examine the effects of home

computers on the social development of children. Two recent exceptions use quasi-experimental

methods to explore the question. Malamud and Pop-Eleches (2010) use a regression

discontinuity design (RDD) to estimate the effects of a government program in Romania that

allocated a fixed number of vouchers for computers to low-income children in public schools.

Their results showed little impact of home computers on children’s engagement in sports or

community service activities. Bauernschuster, Falck and Woessmann (2014) take advantage of a

technological mistake in Internet service provision in Germany to estimate the effects of

broadband access on social capital. Although the focus is on adults, they also examine effects on

children’s extra-curricular school activities such as sports, music, arts, and drama and do not find

evidence of negative effects (Bauernschuster, Falck and Woessmann 2011).

To test the hypothesis that home computers affect social development among children,

we conduct the first-ever randomized control experiment involving the provision of free

computers to children for home use. Half of over one thousand children attending grades 6-10

attending 15 different schools (primarily middle) were randomly selected to receive computers to

use at home. Several measures of social development were tracked over time for the treatment

group of children receiving free computers and the control group of children not receiving free

computers. The experimental design used here solves important problems of selection bias

caused by families choosing to purchase computers for their children. In addition to being the

first study to use a field experiment to explore the question, we also collected a broader range of

measures of social development than prior studies, ranging from in person interactions with

3
friends, to cyberbullying, to school participation. We also focus on adolescence because it is a

sensitive period for the development and maintenance of peer interactions (Crosnoe, 2011) and

thus a crucial time to understand the potentially adverse social developmental impacts of

computer ownership. Earlier findings from the experiment indicate that home computers have no

effect on educational outcomes, such as grades and test scores (Fairlie and Robinson 2013), but

that study does not examine the effects on social development.3

We find that the field experiment has a large effect on computer ownership, total hours of

computer use, and computer use for games, social networking and other entertainment among

children. We find a significant and positive treatment impact on the number of friends children

report communicating with and on the amount of time children report actually hanging out with

their friends (in person). We find no evidence that children randomly assigned to receive a

computer are less likely to participate on sports teams or after-school clubs, or spend less time in

these activities. Finally, treated children are more likely to report using computers for

communication and to have a social networking page. Thus, overall we find evidence of small

positive benefits to children's social development and no statistically significant evidence of

negative effects.

2. Experimental Design

To explore the effects of home computers on social development among children we

conducted a randomized control experiment that provided free personal computers to over one

thousand children for home use. It represents the first field experiment involving the provision of

3
A larger literature explores the impact of home computers on educational outcomes, finding somewhat
mixed results (for a few examples see Attewell & Battle 1999; Fuchs & Woessmann 2004; Fairlie 2005;
Schmitt & Wadsworth 2006; Fiorini, 2010; Malamud & Pop-Eleches 2011; Mo et al. 2012; Fairlie and
Robinson 2013; Vigdor et al. 2014; and see Bulman and Fairlie 2015 for a review).

4
free computers to children for home use ever conducted. The randomized control experiment

removes concerns about selection bias resulting from which families decide to purchase

computers. None of the children participating in the study had computers at baseline. Half were

randomly selected to receive free computers, while the other half served as the control group.

The sample for this study includes 1,123 children enrolled in grades 6-10 in 15 different

middle and high schools in 5 school districts in California. The project took place over two

years: 2008-9 and 2009-10. The 15 schools span the Central Valley of California geographically

and capture both large and small schools, and urban and rural schools. The schools are primarily

middle schools with 97 percent of the sample in grades 6-8. Overall, these schools are similar in

size (749 students compared to 781 students), student to teacher ratio (20.4 to 22.6), and female

to male student ratio (1.02 to 1.05) as California schools as a whole (U.S. Department of

Education 2012). The schools attended by children participating in the experiment, however, are

poorer (81% free or reduced price lunch compared with 57%) and have a higher percentage of

minority students (82% to 73%) than the California average reflecting the necessity of the

experiment to require not having a home computer for eligibility. Participating children also have

lower average test scores than the California average (3.2 compared with 3.6 in English-

Language Arts and 3.1 compared with 3.3 in Math), but the differences are not large (California

Department of Education 2010). Although these differences may impact our ability to generalize

the results, low-income, ethnically diverse schools such as these are the ones most likely to

enroll schoolchildren without home computers and be targeted by policies to address inequalities

in access to technology (e.g. E-rate program and IDAs).

To identify children who did not already have home computers, we conducted an in-class

survey at the beginning of the school year with all of the children attending one of the 15

5
schools. The survey, which took only a few minutes to complete, asked basic questions about

home computer ownership and usage. To encourage honest responses, it was not announced to

children that the survey would be used to determine eligibility for a free home computer (even

most teachers did not know the purpose of the survey). In total, 24 percent of children reported

not having a computer at home. This rate of home computer ownership is roughly comparable to

the national average: – estimates from the Current Population Survey indicate that 27% of

children aged 10-17 do not have a computer with Internet access at home (U.S. Department of

Education 2012).

Any child who reported not having a home computer on an in-class survey was eligible

for the study. All eligible children were given an informational packet, baseline survey, and

consent form to complete at home. To participate, children had to have their parents sign the

consent form and return the completed survey to the school. Of the 1,636 children eligible for the

study, we received 1,123 responses with valid consent forms and completed questionnaires

(68.6%).4 We randomized treatment at the individual level, stratified by school. In total, of the

1,123 participants, 559 were randomly assigned to the treatment group.

In discussing the logistics of the study with school officials, school principals expressed

concern about the fairness of giving computers to a subset of eligible children. For this reason,

we decided to give out computers to all eligible children: treatment children received computers

immediately, while control children had to wait until the end of the school year. Our main

4
This percentage is lowered by two schools in which 35% or fewer children returned a survey because of
administrative problems at the school. However, there may certainly be cases in which children did not
participate because they lost or did not bring home the flier advertising the study, their parents did not
provide consent to be in the study, or they did not want a computer. Thus, participating children are
probably likely to be more interested in receiving computers than non-participating children (which
would also be the case in a real-world voucher or giveaway program). To deal with this, we focus on
Intent-to-Treat effects in our main specifications. Note also that the results we present below are not
sensitive to excluding the two schools with low participation rates.

6
outcomes are all measured at the end of the school year, before the control children received

their computers.

The computers provided through the experiment were purchased from or donated by

Computers for Classrooms, Inc., a Microsoft-certified computer refurbisher located in Chico,

California. The computers were refurbished Pentium machines with 17" monitors, modems,

ethernet cards, CD drives, flash drives, Microsoft Windows, and Microsoft Office (Word, Excel,

PowerPoint, Outlook). The computer came with a 1-year warranty on hardware and software

during which Computers for Classrooms offered to replace any computer not functioning

properly. In total, the retail value of the machines was approximately $400-500 a unit. Since the

focus of the project was to estimate the impacts of home computers on social participation

among children and not to evaluate a more intensive technology policy intervention, no training

or assistance was provided with the computers. We did not directly provide Internet service with

the computers, but hardware was provided and the majority of children receiving free computers

reported subscribing to service.

The computers were handed out by the schools to eligible children in the late fall of the

school year. Almost all of the children sampled for computers received them: we received

reports of only 11 children who did not pick up their computers, and 7 of these had dropped out

of their school by that time. As expected, we found that some of the control group children

purchased home computers by the end of the school year. From a follow-up survey conducted at

the end of the school year, we found that 26 percent of children in the control group purchased

computers, and in most cases these computers were purchased later in the school year (thus

having less potential impacts on measured outcomes). After the distribution, neither the research

team nor Computers for Classrooms had any contact with children during the school year. In

7
addition, all of the outcome measures were collected at least 6 months after the computers were

given out. Thus, it is very unlikely that child behavior would have changed for any reason other

than the computers themselves (for instance, via Hawthorne effects).

Data Sources

Data from the experiment were collected from three main sources. First, we administered

a detailed baseline survey which was required to participate in the project (as that was where

consent was obtained). The survey includes detailed information on child and household

characteristics. Second, each school provided us with detailed administrative data on school

participation for all children covering the entire academic year. Finally, we administered a

follow-up survey at the end of the school year, which included detailed questions about social

development. We collected follow-up surveys for 85 percent of the children. The response rate

was 84.3 percent for the control group and 85.4 percent for the treatment group (the difference is

not statistically significant).5 Administrative data provided by the schools on school participation

has essentially no attrition.

Measures of Social Development

To measure treatment effects on social development we conducted a follow-up survey at

the end of the school year. Information on four broad categories of social development was

included; i) computer-related behavior; ii) social interactions with friends; and iii) participation

in after-school activities, and iv) school participation and engagement. Computer-related social

5
We also do not find evidence that the characteristics of attritors differ between the treatment and control
group. Specifically, we regress an attrition dummy variable on the controls interacted with treatment in
addition to the main variables and do not find any differences.

8
behavior was assessed by asking whether the child ever showed their parent how to do

something on the computer, whether they have a social networking page, whether in the past

school year they had ever been bullied, teased, or threatened online or by email, and whether in

the past school year they had ever received help on school assignments from other students,

friends, or teachers by email, instant messaging or networking.

We designed the questionnaire to include objective measures of social development to the

extent possible. Appendix Table 1 reports the exact wording of each question. Because of the

importance of the effects of home computers on peer interaction we measured this in several

different ways. We asked teenagers about the number of friends they communicate with per

week, the number they hang out with per week (in person), and the number of nights per week

and number of hours per week they hang out with friends (in person). For after-school activities

we asked whether the children participated on a sports team, the number of hours per week the

child spends in sports or exercising, whether the child participates in any school clubs or other

after-school activities, and the number of hours spent participating in clubs/after school

activities.

Our fourth outcome measure is a multi-component objective assessment of school

participation. All of the schools provided us with detailed administrative data on school

participation at the end of the school year. The schools provided information on the number of

absences, unexcused absences, days tardy and days suspended during the school year, and

whether the child was still enrolled at the end of the school year.

9
Summary Statistics and Randomization Verification

Table 1 reports summary statistics for the treatment and control groups and provides a

balance check. In the table, Columns 1 and 2 report the means for the treatment and control

groups, respectively, while Column 3 reports the p-value for a t-test of equality. Demographic

information from the school-provided administrative data and information from the baseline

survey administered to all children is reported. The average age of study participants is 12.9

years. The sample has high concentrations of minority and non-primary English language

students: 55% of students are Latino, and 43% primarily speak English at home. Most students,

however, were born in the United States: the immigrant share is 19%. The average education

level of the highest educated parent is 12.8 years. From the baseline survey, we find that ninety

percent of children live with their mothers, but only 58% live with their fathers. Students report

that 47% of mothers and 72% of fathers are employed (conditional on living with the student).

Overall, we find no differences between the treatment and control groups. None of the

differences are statistically significant or large in magnitude. Nevertheless, we control for all of

these covariates in the regressions that follow.

3. Main Results

A. Computer Ownership, Usage and Activities

The experiment has a very large first-stage impact in terms of increasing computer

ownership and hours of computer use among children. Panel A of Table 2 reports treatment

effects on computer ownership rates and total hours of computer use from the follow-up survey

conducted at the end of the school year. The estimated treatment effects are from linear

regressions that control for school, year, age, gender, ethnicity, grade, parental education,

10
whether the student's primary language is English, whether the student is an immigrant, the

number of people living in the household, whether the parents live with the student, and whether

the parents have a job.6 Treatment effects estimates are also similar without controls. We find

very large effects on computer ownership and usage. We find that 81% of the treatment group

and 26% of the control group report having a computer at follow-up. While this first-stage

treatment effect of 55 percentage points is very large, if anything it is understated because only a

very small fraction of the 559 students in the treatment group did not receive one (as noted

above, we had reports of only 11 children who did not pick up their computer). In addition, any

measurement error in computer ownership would understate the first stage. The treatment group

is also 25 percentage points more likely to report having Internet service at home than the control

group (42% of treatment children have Internet service, compared to 17% of control children).

We also have estimates of total time use on computers from the children. We do not want

to overemphasize these specific estimates of hours use, however, because of potential

measurement error common in self-reported time use estimates. With that caveat in mind, we

find large first-stage results on reported computer usage (also reported in Panel A of Table 2).

The treatment group reports using a computer 2.5 hours more per week than the control group,

which represents a substantial gain over the control group average of 4.2 hours per week. The 4.2

hours that control students spend on computers is spent mostly at school and in other locations

(i.e. libraries, or a friend or relative’s house). But, we do not find evidence of more hours of

computer use by the control group at other locations, including a friend's house, suggesting that

these children did not indirectly benefit from using the computers at the homes of the treatment

6
Some of these variables are missing for some children in the study. To avoid dropping these
observations, we also include a dummy variable equal to 1 if the variable is missing and code the original
variable as a 0 (so that the coefficients are identified from those with non‐missing values). Estimates of
treatment effects are similar dropping these observations.

11
students. Reassuringly, this increase in total hours of computer use comes from home computer

use. The similarity between the point estimate on total computer time and the point estimate on

home computer time suggests that home use does not crowd out computer use at school or other

locations.

The computers were used for both educational and non-educational purposes. In Panel B

of Table 2 we estimate treatment effects on different activities by children. Children spend an

additional 0.8 hours on schoolwork, 0.8 hours per week on games, 0.6 hours on social

networking, and 0.4 hours on email.7 All of these increases are large relative to the control group

means of 1.9, 0.8, 0.6, and 0.3, respectively. Though we do not want to overemphasize the

specific point estimates given possible underreporting of time use, the finding of home computer

use for both schoolwork and entertainment purposes among children is common to numerous

national surveys of computer use (see Pew Internet Project 2008, 2012, 2013; U.S. Department

of Education 2012; Kaiser Family Foundation 2010 for example). The findings are also

suggestive that children report increased use of computers for activities that might be both social

isolating (i.e. game use) and social participating (i.e. email and social networking).

B. ITT vs. LATE

We report the Intent-to-Treat (ITT) estimates of treatment effects from the experiment.

LATE (or IV) estimates would be about twice as large (since the difference in computer usage is

55 percentage points). We do not report these estimates, however, because we cannot technically

scale up the coefficients with the IV estimator because of differential timing of purchasing

computers over the school year by the control group (two thirds of the control group with a home

7
We also find larger medians and distributions that are to the right for the treatment group for these
measures of schoolwork and game/networking use.

12
computer at follow-up obtained this computer after the fall). The finding that 82 percent of the

treatment group reports having a computer at the end of the school year also creates difficulty in

scaling up the ITT estimates because we know that essentially all treatment children picked up

their computers and that many of the treatment group reporting not having a computer at follow-

up indeed had a computer at home (based on subsequent conversations with the children by

principals). For these reasons we focus on the ITT estimates.

C. Treatment Effects on Computer-Related Social Behavior

Table 3 reports estimates of treatment effects on social behavior that is related to

computer use. One clear benefit from having a computer at home is that children are more likely

to report showing their parents how to do something on the computer.8 Having a home computer

might increase communication between children and their parents as the children explain how to

use computers. However, we do not find that home computers increase the likelihood that

children receive help on school assignments from other students, friends, or teachers by email,

instant messaging or social networking.

Having a home computer appears to increase the likelihood of having a social networking

page. The estimated treatment (ITT) effect is sizeable and increases the percentage of the

treatment group having a social networking page by nearly 10 percentage points or roughly 20

percent relative to the control group mean. This is consistent with the increased hours of use of

computers for social networking reported in Table 2. The experiment increases the likelihood of

reporting cyberbullying by 0.03, which is not statistically significant but does nonetheless

8
See Appendix Table 1 for the exact wording of this question and other social development questions on
the follow-up survey.

13
represent almost a 40 percent increase over the (relatively low) control group mean.9 We also do

not find evidence indicating that children are more likely to receive help from a teacher or

classmate through the Internet.

In interpreting these results, one should also keep in mind that many of the treated

children did not establish an Internet connection. Because Internet connectivity was not provided

as part of the experiment and thus not randomly assigned, it is not possible to accurately “scale

up” the results for having a social networking page and experiencing cyberbullying by the share

who actually established an Internet connection (i.e., LATE estimates for those with internet

connections). Nevertheless, if one conducted this thought experiment ignoring likely selection

into obtaining Internet service the impact of home computers with Internet connections on

establishing a social network page and experiencing cyberbullying could be substantially larger.

All of these measures are directly related to social activities on computers. We now turn

to examining measures of social development that are broader or independent of communicating

through computers.

D. Treatment Effects on Social Participation

Home computers could lead to social isolation among children if real-time friendships

and interactions are replaced by virtual ones or if computer use displaces activities that rely on or

promote social engagement. Even with one activity such as gaming for example, children may

become isolated playing games directly against the computer, but also get together to play the

same games or interactive games (Orleans and Laney 2000).

9
Cyberbullying has received a considerable amount of attention recently. See, for example, "How Big a
Problem Is Bullying or Cyberbullying in Your School or Community?" NY Times, September 18, 2013,
and "Experts Offer Their Guidance for Dealing With Online Bullying" NY Times, March 24, 2014.

14
 We explore this question by examining the treatment effects on children's reports of

social interactions with friends and participation in after-school clubs and activities. Table 4

reports estimates. We find that home computers actually increase the number of friends that

children communicate with outside of school hours and increase the number of hours hanging

out with friends (in person). Both of the estimated treatment effects are large relative to the

control group mean. The treatment group communicates with an average of 1.57 more friends

per work and hangs out for 0.72 more hours per week than the control group. For the other

measures of social interactions with friends we find positive, but statistically insignificant point

estimates.

To provide additional evidence on the overall effects of home computers on social

participation we create a summary index that aggregates information over multiple treatment

effect estimates (Kling, Liebman, & Katz 2007). Specifically, we create an index of the social

interaction with friends measures that combines the four measures reported in Table 4. By

aggregating the separate educational outcomes we improve the statistical power of treatment

effect estimates. To create the index we first calculate z-scores for each of the dependent

variables by subtracting the control group mean and dividing by the control group standard

deviation. Thus, each dependent variable has mean zero and standard deviation equal to one for

the control group. The educational outcome index is then calculated from an equally-weighted

average of the z-scores for the four dependent variables. The treatment effect estimate for this

index indicates where the mean of the treatment group is in the distribution of the control group

in terms of standard deviation units.

Appendix Table 2 reports estimates for the social interaction with friends outcome index.

By definition the control group mean for the index is 0. The treatment effect estimate is positive

15
and statistically significant. The point estimate of 0.10 implies that the treatment group mean is

0.10 standard deviations higher than the control group mean. The treatment effect estimate for

the summary measure of educational outcomes provides additional evidence that home

computers increase social interactions with friends.

In Table 5, we examine the effects of home computers on another measure of social

development by examining participation in sports, clubs, and other after-school activities.

Participation in after-school activities provides an important and objective measure of social

development. Being part of a sports team, club, or music group is hypothesized to promote

children’s intellectual and emotional adjustment in part because doing so provides an organized

forum for positive peer engagement, thereby generating social and human capital (Eccles &

Barber, 1999; Eccles & Templeton, 2002; Feldman & Matjasko, 2005; Mahoney et al. 2009).

Using measures of after-school sports and club participation and hours, we find no evidence of

negative effects of home computers. Estimates for a summary z-score for these four variables

confirm this finding (see Appendix Table 2). Having a home computer does not appear to lead to

increasing social isolation by crowding out after-school activities.

E. Treatment Effects on School Participation and Engagement

We also measure whether home computers affect participation and engagement in school

through their impacts on the number of absences, unexcused absences, tardies and days

suspended, and whether the child was still enrolled in the school at the end of the school year.

Any negative social impacts of home computers might spill over into how schoolchildren

participate and engage in school. We measure each of the outcomes as both the number of events

(i.e. times or days) and incidence (i.e. 0/1). Table 6 reports estimates for each measure, and

16
Appendix Table 2 reports the summary z-score estimate. We find no evidence of negative effects

of home computers on school participation and engagement.

4. Treatment Heterogeneity

In this section, we explore heterogeneity in treatment effects by various baseline

characteristics. We focus specifically on pre-treatment measures of social participation, social

communication, and parental controls. Table 7 reports estimates of heterogeneous treatment

effects for prior social participation and parental controls. We first report estimates for

interacting treatment with children reporting hanging out with many friends at baseline (defined

as 4 or more which is roughly the median). We do not find evidence of differential treatment

effects by baseline interaction with hanging out with many friends. We also report estimates for

interacting treatment with children reporting that they have a curfew at baseline. We find a

couple of positive treatment interaction point estimates with one being statistically significant,

but the results are not strong. It might be that the computer enabled children who have a curfew

to reach out to more friends because they can now communicate with those friends more easily

when they are at home.

In Table 8, we focus on measures of prior social communication. On the baseline survey

we asked children if they had a social networking page (which 40 percent did) and whether they

sent text messages (which 35 percent did). We interact treatment with both of these measures of

pre-treatment social communication. We find some evidence suggesting a positive relative effect

on social participation outcomes for students who did not use these forms of communicating with

friends at baseline. The home computers allowed these children who were previously socially

participating at lower levels to catch up, or at least partially catch up, with children who were

17
previously socially participating at higher levels (note the interaction treatment effects are large

relative to the negative main effects). Thus, rather than increasing social isolation as has been

feared by some, access to home computers appears to play a compensatory role for those

students who were initially relatively less socially connected in the digital world.

We also examined treatment heterogeneity by gender, ethnicity and grade level.

Appendix Table 3 reports estimates. We do not find statistically significant treatment interactions

across these demographic groups.

5. Conclusion

Do computers inhibit children’s social development and lead to social isolation, or do

they facilitate in-person interactions with friends and schoolmates? There is surprisingly little

credible research addressing this question. We provide the first relevant evidence using a

randomized control experiment. The field experiment involved the random provision of

computers to use at home to more than one thousand children attending 15 different schools. We

found that home computers increase total use of computers for social networking and email

among children as well as the total use of computers for games and other entertainment. Far from

being socially isolating, however, we find a significant and positive impact on the number of

friends treated children report communicating with and on the amount of time children report

actually hanging out with their friends in person. We also find no evidence that children

randomly assigned to receive a computer are less likely to participate in sports teams or after-

school clubs, or spend any less time in these activities. All told, the results portray a pattern of

small positive benefits to youth’s social development and no significant evidence of increasing

social isolation, at least according to the outcomes we have measured here.

18
 The underlying mechanisms that give rise to these effects merit further exploration.

Children might spend time alone at home on the computer, but computers may provide

opportunities for communication with others and a focus point for friends to gather around.

Increased computer time also does not appear to displace other social activities, but this requires

further study. Clearly, a treatment-induced increase in time spent on the computer and hanging

out with friends must be offset by a reduction in time spent on something else. Our follow-up

survey provides information on TV time, homework time, and reading (non-homework) time,

but we did not find negative (or positive) treatment-control differences for any of these variables.

Indeed, the treatment point estimates were all very close to zero with or without controls. The

time could be coming from activities that we did not measure such as sleeping, time with family,

eating, and exercising, but we cannot be sure. Further research in this vein could consider

including a time diary that forces a summing up of time across activities and results in more

accurate measures of time use along with including more options. However, such time diaries are

difficult to administer, time consuming, and often experience low response rates among children.

Additional measures of social development may also be important and merit further

study. Moreover, home computers may increase social and academic development by increasing

home-school connections. One clear benefit from having a computer at home is that children are

more likely to report showing their parents how to do something on the computer. We also find

some evidence from our experiment of a positive effect of home computers on whether parents

check assignments, grades and attendance online using student information system software.

(results not reported here). These types of software, which provide parents with nearly

instantaneous information on their children's school performance, attendance and disciplinary

actions, are becoming increasingly popular in schools. More research is needed on how access to

19
home computers affects home-school connections such as these and how they ultimately affect

children's social and academic development.

20
References

Attewell, P. & Battle, J. (1999). Home Computers and School Performance. The Information
Society 15: 1-10.

Bauernschuster, Stefan, Oliver Falck, and Ludger Woessmann. (2014) Surfing alone? The
Internet and social capital: Evidence from an unforeseeable technological mistake.
Journal of Public Economics 117: 73-89.

Bauernschuster, Stefan, Oliver Falck, and Ludger Woessmann. (2011). Surfing Alone? The
Internet and Social Capital: Evidence from an Unforeseeable Technological Mistake.
CESifo Working Paper 3469. CESifo, Munich.

Bulman, George, and Robert W. Fairlie. (2015). "Technology and Education: Computers,
Software, and the Internet," Handbook of the Economics of Education, Vol. 5, eds Eric
Hanushek, Steve Machin, and Ludger Woessmann, North-Holland, Chapter 6: 239-280.

California Department of Education (2010). 2010 STAR Test Results: California STAR
Program, http://star.cde.ca.gov/star2010/

Crosnoe, R. (2011). Fitting In, Standing Out: Navigating the Social Challenges of High School
to Get an Education. New York: Cambridge University Press.

Eccles, J. S., & Barber, B. L. (1999). Student council, volunteering, basketball, or marching
band: What kind of extracurricular involvement matters? Journal of Adolescent
Research, 14, 10–43.

Eccles, J. S., & Templeton, J. (2002). Extracurricular and other after-school activities for youth.
Review of Research in Education, 26, 113–180.

Fairlie, Robert W. (2004). "Race and the Digital Divide," Contributions to Economic Analysis &
Policy, The Berkeley Electronic Journals 3(1), Article 15: 1-38.

Fairlie, R. (2005). The Effects of Home Computers on School Enrollment. Economics of

Education Review 24, 533-547.

Fairlie, R., & Robinson, R. (2013). Experimental Evidence on the Effects of Home Computers
on Academic Achievement among Schoolchildren. American Economic Journal: Applied
Economics, 5, 211-40.

Feldman, A. & Matjasko, J. (2005). The role of school-based extracurricular activities in

adolescent development: A comprehensive review and future directions. Review of
Educational Research, 75, 159-210.

Fiorini, M. (2010). The Effect of Home Computer Use on Children’s Cognitive and Non-
Cognitive Skills. Economics of Education Review 29, 55-72.

21
Fuchs, T., & Woessmann, L. (2004). Computers and Student Learning: Bivariate and
Multivariate Evidence on the Availability and Use of Computers at Home and at School.
CESifo Working Paper No. 1321.

Goldfarb, A. & Prince, J. (2008). Internet Adoption and Usage Patterns are Different:
Implications for the Digital Divide. Information Economics and Policy 20, 2-15.

Kaiser Family Foundation. (2010). Generation M2: Media in the Lives of 8- to 18-Year Olds.
Kaiser Family Foundation Study.

Kling, J., Liebman, J., & Katz, L. (2007). Experimental Analysis of Neighborhood Effects.
Econometrica 75, 83–119.

Kraut, R., Patterson, M., Lundmark, V., Kiesler, S., Mukopadhyay, T., & Scherlis, W. (1998).
Internet paradox: A social technology that reduces social involvement and psychological
well-being? American Psychologist, 53, 1017-1031.

Mahoney, J. L., Vandell, D. L., Simpkins, S., & Zarrett, N. (2009). Adolescent out-of-school
activities. In R. M. Lerner & L. Steinberg (Eds.), Handbook of adolescent psychology:
Vol. 2. Contextual influ- ences on adolescent development (3rd ed., pp. 228 –269).
Hoboken, NJ: Wiley.

Malamud, O. & Pop-Eleches, C. (2011). Home Computer Use and the Development of Human
Capital. Quarterly Journal of Economics 126, 987-1027.

Mo, D., Swinnen, J., Zhang, L., Yi, H., Qu, Q., Boswell, M., and Rozelle, S. 2012. "Can One
Laptop per Child Reduce the Digital Divide and Educational Gap? Evidence from a
Randomized Experiment in Migrant Schools in Beijing," Rural Education Action Project,
Stanford University, Working Paper 233.

Mossberger, K., Tolbert, C., & Stansbury, M. (2003). Virtual Inequality: Beyond the Digital
Divide. Georgetown University Press, Washington, DC.

National Telecommunications and Information Administration. (2013). Exploring the Digital

Nation: America’s Emerging Online Experience.

Ono, H. & Zavodny, M. (2007). Digital Inequality: A Five Country Comparison Using
Microdata. Social Science Research 36, 1135-1155.

Orleans, M., and M. Laney. 2000. "Children’s Computer Use in the Home: Isolation or
Socialization?,"
Social Science Computer Review. 18(1): 56-72.

22
Pew Internet Project. 2008. Teens, Video Games, and Civics: Teens’ gaming experiences are
diverse and include significant social interaction and civic engagement, Washington,
D.C.: Pew Internet & American Life Project.

Pew Internet Project. 2012. How Teens Do Research in the Digital World, Washington, D.C.:
Pew Internet & American Life Project, http://www.pewinternet.org/2012/11/01/how-
teens-do-research-in-the-digital-world/

Pew Internet Project. 2013. Teens, Social Media, and Privacy, Washington, D.C.: Pew Internet
& American Life Project.

Schaefer, D., Simpkins, S., Vest, A., & Price, C. (2011). The contribution of extracurricular
activities to adolescent friendships: New insights through social network analysis.
Developmental Psychology, 47, 1141-1152.

Servon, Lisa. 2002. Bridging the Digital Divide: Community, Technology and Policy
(Blackwell).

Schmitt, J. & Wadsworth, J. (2006). Is There an Impact of Household Computer Ownership on

Children's Educational Attainment in Britain? Economics of Education Review, 25, 659-
673.

Subrahmanyam, K., Greenfield, P., Kraut, R., & Gross, E. (2001). The impact of computer use
on children's and adolescents' development. Applied Developmental Psychology, 22, 7-
30.

U.S. Department of Education (2012). Digest of Education Statistics 2012. National Center for
Education Statistics, Institute of Education Sciences, U.S. Department of Education.
Washington, DC.

Vigdor, Jacob L., Helen F. Ladd, and Erika Martinez. 2014. “Scaling the Digital Divide: Home
Computer Technology and Student Achievement,” Economic Inquiry. 52(3): 1103–1119.

Zavodny, M. (2006). Does Watching Television Rot Your Mind? Estimates of the Effect on Test
Scores. Economics of Education Review 25, 565-573.

23
 Appendix Table 1: Follow-Up Survey Questions for Measures of Social Development

A. Social development related to computer use

A1. Do you ever show your parents how to do something on the computer?
A2. Do you have a social networking page (MySpace, Facebook or Hi5 page)?
A3. Over the past school year have you ever been bullied, teased, or threatened online or by
email?
A4. Over the past school year have you ever received help on school assignments from other
students, friends, or teachers by email, instant messaging or networking?

B. Social development related to interaction with friends

B1. In a normal week, about how many friends do you communicate with (phone, email, instant
messaging, etc...) after school, at night, or on the weekends?
B2. In a normal week, about how many friends do you hang out with after school, at night, or on
the weekends?
B3. In a normal week (7 days), how many nights do you hang out with friends outside of your
home or someone else's home?
B4. How many hours do you spend per week in the following activities? hanging out with friends
(at malls, movies, houses, etc.)

C. Social development related to sports, clubs and other after-school activities

C1. Do you participate on a sports team?
C2. How many hours do you spend per week in the following activities? playing sports or
exercising
C3. Do you participate in school clubs or any other after-school activities?
C4. How many hours do you spend per week in the following activities? In other after-school
activities (clubs, Boy/Girl Scouts, etc.)

24
Appendix Table 2
Z‐Scores for Summary Measures of Social Participation

Sports, clubs and School All three

Hanging out with
other activities participation categories
friends summary
summary summary summary

Treatment effect 0.10 0.01 ‐0.03 0.01
(0.05)* (0.05) (0.03) (0.03)
Observations 636 826 1044 586
Control mean ‐0.05 0.00 0.01 ‐0.04
Control std. dev. 0.65 0.66 0.62 0.38
Notes. OLS regressions are estimated for each specification. Heteroscedasticity‐consistent standard errors
are reported in parentheses. To create the indices we first calculate z‐scores for each of the underlying
dependent variables by subtracting the control group mean and dividing by the control group standard
deviation. The index is then calculated from an equally‐weighted average of the z‐scores for the underlying
dependent variables.
***, **, * indicates significance at 1, 5 and 10%.

25
Appendix Table 3
Heterogeneity by Baseline Demographic Characteristics
(1) (2) (3) (4) (5) (6) (7)
Hours hanging
Friends Friends Sports, clubs School
Hours of Bullied online out with
communicate summary z‐ and activities participation z
computer use or via email friends per
with per week score z‐score score
week
Panel A. Female Interaction
Treatment 2.01 0.03 2.25 0.92 0.10 ‐0.02 ‐0.01
(0.68)*** (0.03) (0.89)** (0.57) (0.08) (0.07) (0.05)
Female student ‐0.69 ‐0.01 0.33 ‐0.49 ‐0.16 ‐0.33 ‐0.10
(0.58) (0.03) (0.82) (0.50) (0.07)** (0.06)*** (0.05)**
Female student * treatment 1.04 0.00 ‐1.74 ‐0.53 ‐0.01 0.07 ‐0.04
(0.99) (0.04) (1.17) (0.73) (0.11) (0.10) (0.06)
Observations 755 852 820 847 636 826 1044

Panel B. Latino Interaction

Treatment 2.41 0.02 1.10 1.15 0.12 0.08 ‐0.05
(0.67)*** (0.03) (0.97) (0.73) (0.10) (0.07) (0.05)
Latino 0.03 ‐0.17 2.25 ‐0.68 ‐0.05 ‐0.04 0.11
(1.04) (0.12) (1.61) (1.96) (0.22) (0.17) (0.09)
Latino * treatment 0.20 0.03 0.50 ‐0.91 ‐0.04 ‐0.11 0.04
(0.91) (0.04) (1.20) (0.84) (0.12) (0.10) (0.07)
Observations 755 852 820 847 636 826 1044

Panel C. Grade Interactions

Treatment 2.44 0.01 1.20 1.27 0.10 0.06 ‐0.05
(0.64)*** (0.03) (0.89) (0.58)** (0.07) (0.06) (0.04)
Grade 6 * treatment 0.39 0.07 0.11 ‐1.88 ‐0.07 ‐0.09 ‐0.13
(1.80) (0.07) (1.84) (1.12)* (0.18) (0.16) (0.13)
Grade 8+ * treatment 0.11 0.03 0.37 ‐1.05 0.01 ‐0.08 0.06
(1.02) (0.04) (1.27) (0.78) (0.11) (0.10) (0.07)
Observations 755 852 820 847 636 826 1044
Notes: OLS or LPM regressions are estimated for each specification. Heteroscedasticity‐consistent standard errors are reported in parentheses.
Dependent variables are from a follow‐up survey completed by students. See notes to Table 2.
***, **, * indicates significance at 1, 5 and 10%.

26
Table 1
Individual Level Summary Statistics and Balance Check
Equality of
Control Treatment means p‐
value
Student and family characteristics
Age 12.91 12.90 0.91
Female 0.51 0.50 0.66
Ethnicity = African American 0.13 0.14 0.86
Ethnicity = Latino 0.56 0.55 0.76
Ethnicity = Asian 0.12 0.14 0.42
Ethnicity = White 0.16 0.14 0.56
Immigrant 0.21 0.18 0.15
Primary language is English 0.43 0.43 0.97
Parent's education 12.81 12.76 0.64
Number of people living in household 4.98 5.02 0.79
Lives with mother 0.92 0.89 0.12
Lives with father 0.58 0.58 0.90
Does your mother have job? 0.47 0.46 0.68
Does your father have a job? 0.73 0.70 0.36
Notes. N=1,123. Columns 1 and 2 report means. Column 3 reports the p‐value for the t‐test for the
equality of means.
***, **, * indicates significance at 1, 5 and 10%.

27
 Table 2
Effect of Program on Computer Ownership and Usage
Hours of Computer Use Per Week
Owns a Has Internet
At Other
Computer Connection Total At Home At School
Location

Panel A. Computer Ownership and
Usage

Treatment 0.55 0.24 2.52 2.59 ‐0.02 ‐0.05
(0.03)*** (0.03)*** (0.48)*** (0.31)*** (0.17) (0.30)
Observations 852 831 755 755 755 755
Control mean 0.26 0.17 4.23 0.76 1.59 1.89
Control std. dev. 0.44 0.38 5.22 2.31 2.32 3.98

Hours of Computer Use Per Week
Net‐
Schoolwork Email Games Other
working
Panel B. Activities on Computer

Treatment 0.77 0.42 0.83 0.59 0.18
(0.25)*** (0.12)*** (0.24)*** (0.18)*** (0.11)
Observations 671 671 671 671 671
Control mean 1.89 0.25 0.84 0.57 0.62
Control std. dev. 2.57 0.72 1.81 1.79 1.39
Notes: OLS or LPM regressions are estimated for each specification. Heteroscedasticity‐consistent standard
errors are reported in parentheses. Dependent variables are from a follow‐up survey completed by students.
Regressions control for the sampling strata (school*year) and baseline characteristics (age, gender, ethnicity,
grade, parental education, whether the student's primary language is English, whether the student is an
immigrant, number of people living in household, whether the mother/father lives with the student, and whether
the mother/father has a job). To avoid dropping observations, for each variable, we create a dummy equal to 1 if
the variable is missing for a student and code the original variable as a 0 (so that the coefficients are identified
from those with non‐missing values).

***, **, * indicates significance at 1, 5 and 10%.

28
Table 3
Social Development Related to Computer Use

Show parents Do you have a Received help
how to do social Bullied online or from teacher or
something on networking via email classmate via
computer page? Internet / email

Treatment effect 0.43 0.09 0.03 0.02
(0.03)*** (0.04)** (0.02) (0.03)
Observations 700 692 852 851
Control mean 0.12 0.53 0.08 0.37
Control std. dev. 0.32 0.50 0.28 0.48
Notes. LPM regressions are estimated for each specification. Heteroscedasticity‐consistent standard errors
are reported in parentheses. Dependent variables are from a follow‐up survey completed by students.
Regressions control for the sampling strata (school*year) and baseline characteristics (age, gender,
ethnicity, grade, parental education, whether the student's primary language is English, whether the
student is an immigrant, number of people living in household, whether the mother/father lives with the
student, and whether the mother/father has a job). To avoid dropping observations, for each variable, we
create a dummy equal to 1 if the variable is missing for a student and code the original variable as a 0 (so
that the coefficients are identified from those with non‐missing values).
***, **, * indicates significance at 1, 5 and 10%.

29
Table 4
Social Development Related to Interactions with Friends
Hours per week
Friends Hours per week
Friends hang out Nights hang out phone and
communicate hanging out with
with outside outside texting with
with per week friends
friends

Treatment effect 1.57 0.31 0.22 0.72 0.03
(0.59)*** (0.45) (0.16) (0.39)* (0.49)
Observations 820 837 679 847 846
Control mean 6.71 5.62 1.72 3.16 3.38
Control std. dev. 8.19 6.52 1.89 5.16 7.19
Notes. OLS regressions are estimated for each specification. Heteroscedasticity‐consistent standard errors are reported
in parentheses. Dependent variables are from a follow‐up survey completed by students. Regressions control for the
sampling strata (school*year) and baseline characteristics (age, gender, ethnicity, grade, parental education, whether
the student's primary language is English, whether the student is an immigrant, number of people living in household,
whether the mother/father lives with the student, and whether the mother/father has a job). To avoid dropping
observations, for each variable, we create a dummy equal to 1 if the variable is missing for a student and code the
original variable as a 0 (so that the coefficients are identified from those with non‐missing values).
***, **, * indicates significance at 1, 5 and 10%.

30
Table 5
Social Development Related to Sports, Clubs and Other After‐School Activities
School
Hours per week
Sports team Hours per week club/after‐
in after‐school
participation playing sports school activity
activities
participation

Treatment effect ‐0.01 0.42 ‐0.03 0.02
(0.03) (0.28) (0.03) (0.16)
Observations 850 849 846 844
Control mean 0.36 2.59 0.38 0.92
Control std. dev. 0.48 3.28 0.49 2.22
Notes. OLS or LPM regressions are estimated for each specification. Heteroscedasticity‐consistent standard
errors are reported in parentheses. Dependent variables are from a follow‐up survey completed by
students. Regressions control for the sampling strata (school*year) and baseline characteristics (age,
gender, ethnicity, grade, parental education, whether the student's primary language is English, whether
the student is an immigrant, number of people living in household, whether the mother/father lives with
the student, and whether the mother/father has a job). To avoid dropping observations, for each variable,
we create a dummy equal to 1 if the variable is missing for a student and code the original variable as a 0
(so that the coefficients are identified from those with non‐missing values).
***, **, * indicates significance at 1, 5 and 10%.

31
Table 6
Social Development Related to School Participation and Engagement

Unexcused Number of Days Still enrolled at

Total Absences
Absences Tardies Suspended End of Year

Panel A. Number of Times
Treatment effect ‐0.63 ‐0.33 ‐0.26 ‐0.30
(0.61) (0.39) (0.93) (0.29)
Observations 1044 1104 1104 1106
Control mean 10.81 4.94 11.53 1.41
Control std. dev. 11.87 7.84 17.00 6.50

Panel B. Percentage with at Least One Occurrence
Treatment effect 0.00 0.01 ‐0.03 0.00 0.01
(0.02) (0.02) (0.02) (0.02) (0.02)
Observations 1044 1104 1104 1106 1123
Control mean 0.86 0.65 0.87 0.17 0.88
Control std. dev. 0.35 0.48 0.34 0.37 0.33
Notes. OLS or LPM regressions are estimated for each specification. Heteroscedasticity‐consistent standard
errors are reported in parentheses. Dependent variables are from school‐provided administrative data.
Regressions control for the sampling strata (school*year) and baseline characteristics (age, gender, ethnicity,
grade, parental education, whether the student's primary language is English, whether the student is an
immigrant, number of people living in household, whether the mother/father lives with the student, and
whether the mother/father has a job). To avoid dropping observations, for each variable, we create a dummy
equal to 1 if the variable is missing for a student and code the original variable as a 0 (so that the coefficients
are identified from those with non‐missing values).
***, **, * indicates significance at 1, 5 and 10%.

32
Table 7
Heterogeneity by Baseline Social Participation and Parental Control
Hours
Friends Friends Sports, clubs School
Hours of Bullied online hanging out
communicate summary z‐ and activities participation
computer use or via email with friends
with per week score z‐score z‐score
per week
Panel A. Hang out with Many Friends at Baseline
Treatment effect 2.13 0.04 1.15 0.24 0.12 0.01 ‐0.03
(0.64)*** (0.03) (0.75) (0.53) (0.07)* (0.06) (0.04)
Hang out with many friends 0.38 0.01 3.12 0.70 0.34 0.15 0.11
at baseline (0.55) (0.03) (0.83)*** (0.54) (0.07)*** (0.06)** (0.05)**
0.82 0.00 0.48 0.93 ‐0.02 0.01 0.00
Many friends * treatment
(1.00) (0.04) (1.18) (0.81) (0.11) (0.09) (0.06)
Observations 720 815 787 811 621 794 1003

Panel B. Parents have Curfew at Baseline

Treatment effect 4.07 0.05 0.64 ‐0.84 0.02 0.09 ‐0.09
(1.32)*** (0.05) (1.39) (0.69) (0.11) (0.11) (0.07)
0.01 ‐0.01 ‐1.31 ‐1.54 ‐0.10 0.06 ‐0.03
Have curfew at baseline
(0.81) (0.04) (1.19) (0.54)*** (0.10) (0.08) (0.06)
‐1.63 ‐0.01 1.09 1.84 0.09 ‐0.07 0.09
Curfew * treatment
(1.40) (0.05) (1.52) (0.81)** (0.13) (0.12) (0.08)
Observations 723 813 784 808 606 788 998
Notes: OLS or LPM regressions are estimated for each specification. Heteroscedasticity‐consistent standard errors are reported in
parentheses. Dependent variables are from a follow‐up survey completed by students. See notes to Table 2.
***, **, * indicates significance at 1, 5 and 10%.

33
Table 8
Heterogeneity by Baseline Social Communication
Hours
Friends Friends Sports, clubs School
Hours of Bullied online hanging out
communicate summary z‐ and activities participation
computer use or via email with friends
with per week score z‐score z‐score
per week
Panel A. No Social Networking Page at Baseline
Treatment effect 3.40 ‐0.01 ‐0.68 ‐0.50 ‐0.09 ‐0.10 ‐0.04
(0.71)*** (0.04) (1.05) (0.61) (0.09) (0.07) (0.06)
No social networking 0.53 ‐0.07 ‐4.44 ‐1.51 ‐0.31 ‐0.06 ‐0.11
page at baseline (0.50) (0.03)** (0.94)*** (0.58)*** (0.08)*** (0.07) (0.05)**
No social networking ‐1.20 0.06 3.02 1.84 0.28 0.18 0.01
* treatment (0.95) (0.04) (1.25)** (0.78)** (0.11)** (0.10)* (0.07)
Observations 743 840 811 835 627 814 1032

Panel B. No Text Messages at Baseline

Treatment effect 2.87 0.01 1.07 ‐0.21 0.03 ‐0.13 ‐0.03
(0.87)*** (0.04) (1.12) (0.79) (0.09) (0.08) (0.06)
No text messages at 0.32 ‐0.04 ‐3.94 ‐1.58 ‐0.27 ‐0.15 ‐0.07
baseline (0.54) (0.03) (0.96)*** (0.60)*** (0.08)*** (0.07)** (0.05)
No text messages * ‐0.54 0.05 0.62 1.37 0.11 0.18 ‐0.03
treatment (1.06) (0.05) (1.32) (0.91) (0.11) (0.10)* (0.07)
Observations 717 806 783 802 605 783 988
Notes: OLS or LPM regressions are estimated for each specification. Heteroscedasticity‐consistent standard errors are reported in
parentheses. Dependent variables are from a follow‐up survey completed by students. See notes to Table 2.
***, **, * indicates significance at 1, 5 and 10%.

34